Systeme, vorrichtungen und verfahren zur anomaliedetektion
Patent Information
- Application Number
- AT2021743348T
- Authority / Receiving Office
- AT · AT
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-06-25
- Filing Date
- 2021-06-24
- Publication Date
- 2026-03-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In large-scale distributed systems and cloud environments, existing anomaly detection methods face challenges such as ill-defined notions of anomalies, lack of labeled data, scalability issues, and the need to handle diverse time-series data with varying patterns, making it difficult to detect anomalies efficiently and accurately.
The approach models time series of probability distributions over real values, allowing for unsupervised anomaly detection in streaming data, using a system with an anomaly detection service that ingests data, applies machine learning models, and generates alerts and recommendations, while considering the entire distribution of measurements within each time unit, and employing a recurrent neural network for prediction.
This method enables efficient detection of anomalies in near real-time, reducing application disruptions, improving remediation time, and enabling engineers to focus on innovative work by automatically detecting operational incidents and identifying operational risks.